Linguistic commentary from a guy who takes things too literally

Songs We Have to Play or the Fans Get Angry

Posted by Neal on September 15, 2012

If you read an interview with some classic rock act that still goes on tour, sometimes they’ll talk about their greatest hits that the audiences always expect to hear, and say something like

There are songs [we have to play ___] or [the fans will get angry].

Sentences like that one interest me, because they appear to be one variety of a well-studied family of unusual coordinate structures, but somehow this variety hasn’t been fit into the family picture yet. Before I go further, I’d better introduce the family for newer readers, or re-introduce it for longtime readers who might not remember the details.

In several posts, I’ve written about so-called non-ATB extractions. First off, extraction refers to constructions in which something (whether a noun phrase, or an adverb, or something else) is missing from a clause in the place where you’d expect it, and instead, either it or something like a relative pronoun or interrogative word is placed at the beginning of the clause. As hinted in this definition, typical examples include relative clauses (the meal I cooked ___) and wh-questions (Where did you say you put it ___?). Other examples include topicalization (Him, I never liked ___). The missing part is sometimes called the gap, and filler refers to the relative pronoun, interrogative word or phrase, or displaced phrase that’s understood to fill that gap.

Moving on, ATB stands for across-the-board, and refers to extractions from coordinated clauses or verb phrases in which a single filler corresponds to a gap in every clause or verb phrase. For example:

things you [hear ___] or [read ___]

Non-ATB extractions, then, are extractions involving a coordination of clauses or verb phrases, some of which contain a gap and some of which don’t, and a single filler for those gaps. This kind of coordination isn’t always grammatical. For it to work, one of several kinds of relationship has to hold between the coordinated clauses or verb phrases. In the literature so far, three have been identified:

  1. Occasion
    In this kind of example, the clauses or VPs refer to a sequence of events that naturally go together to talk about something that can be considered a single event (albeit composed of several parts). The first example of this kind that I blogged about was from the song “Neon Moon” by Brooks & Dunn, which contained the line

    No telling how many tears I’ve [sat here] and [cried ___].

  2. Result
    In this kind, the clause or VP with a gap refers to something that causes the event that the other clause or VP refers to. Here’s an example that I blogged about from a comic strip:

    The perfect course for me to [enroll in ___] and [meet new people]!

  3. Violated Expectation
    In this kind, the clause or VP with a gap refers to something that you would think would cause the opposite of the clause without a gap, but surprise! It didn’t. Classic examples include questions like

    How much can you [drink __] and [not get drunk]?

  4. The specific names for these relations come from Andy Kehler of UCSD, who classifies each of them as a subtype of one of three coherence relations (a term he takes from the philosopher David Hume). The three coherence relations are Resemblance, Contiguity, and Cause/Effect. The subtypes of relationships are all the usual relations between thoughts that you’re accustomed to: exemplification, explanation, contrast, etc. Some of these relations, as it happens, can be expressed by the conjunction and, which is where these relations become relevant for extractions out of coordinate structures. Ordinary ATB extractions are examples of the relation Resemblance, and more specifically, a relation that Kehler calls Parallel. Occasion is a subtype of the Contiguity relation. Both Result and Violated Expectation are subtypes of Cause/Effect.

    All these examples of non-ATB extractions have involved and. Some, however, involve or, like the one at the beginning of this post, and others here and here.

    Interestingly, all the or examples I’ve found have involved some kind of obligation-related meaning. I’ve wondered why for quite some time.

    Meanwhile, in another corner of English syntax, there are constructions variously known as pseudo-imperatives, conditional imperatives, or (for reasons which will become clear) Imperative-and-Declaratives (IaDs for short). They’ve also been given the longer name of syntactic coordination despite semantic subordination by Peter Culicover and Ray Jackendoff in a 1997 paper and in their 2005 book Simpler Syntax. They consist of sentences like these:

    Say the word, and it’s yours.

    Culicover and Jackendoff noted a similarity between these IaDs and non-ATB extractions. Specifically, when you translate an IaD into an extraction involving a coordination, it has to be a non-ATB extraction. Try to make it ATB, and you change the meaning. (See what I did there?) C&J give this example:

    Identify the thief, and we’ll arrest him on the spot.
    ?This is the thief that [you identify ___] and we’ll arrest him on the spot. (non-ATB. Awkward, but if you do accept it, it has conditional meaning: “If you identify him, we’ll arrest him.”)
    ?This is the thief that [you identify ___] and [we’ll arrest ___] on the spot. (ATB. Also awkward, but if you do accept it, there’s no conditional meaning. It means that both of these things are supposed to happen.)

    Now as it turns out, there are also pseudo-imperatives with or — that is, Imperative-or-Declaratives (IoDs). These don’t say what will happen if a condition is fulfilled; they say what will happen if a condition is not fulfilled. Usually, these are undesirable situations, so they’re basically threats, and C&J give this usage of or the name Threat-Or. For example:

    Let me go, or you’ll be sorry.

    C&J talk about these, too, and observe that like IaDs, they have an analogous form as a non-ATB extraction. They give this example:

    Take that linguist seriously or you risk your career.
    That is one linguist that [you take ___ seriously] or [you risk your career].

    I’d forgotten about examples like this from when I read their 1997 paper, and it only occurred to me recently that my non-ATB examples with or were a lot like these pseudo-imperatives, and that C&J just might have mentioned them along with their and examples. Except that their examples are made up, and mine are actual utterances.

    So now I want to connect with C&J have said about Threat-Or/IoDs with what Kehler has said about non-ATB extractions. I want to propose a third type of Cause/Effect relation, which you might call Result to Be Avoided (to stand alongside Unexpected Result), or which you could just call Threat (following C&J).

    The reason that non-ATB extractions with or all seem to involve an idea of necessity is that in the coherence relation of Threat is always expressing one even that must happen in order to prevent another one. In non-extraction constructions, the way to express necessity is with an imperative (Stop or I’ll shoot) or a deontic modal (You {had better, have to, should,…} stop, or I’ll shoot). Of these two means of expressing necessity, only modals are available in extraction constructions.

    The required meaning of necessity also explains why we don’t get the Threat reading in sentences like

    He stopped, or I shot.

    After the events have taken place, there’s no threat anymore. Furthermore, if there ever was a threat, then pragmatically you’d expect the speaker at this point to say whether the threat was avoided or carried out, not to be coy and say, “Either it happened, or it didn’t.”

    For those who are interested, here are my paraphrases of Kehler’s formal definitions of the Cause/Effect relations of Result and Violated Expectation (from Chapter 2 of Coherence, Reference, and the Theory of Grammar), along with my proposal for Threat. In these definitions, P ~~> Q is a less strict version of implication, meaning “Q could plausibly follow from P”.

    • Result: For the utterance of Sentence1 with meaning P and Sentence2 with meaning Q to be coherent, it must be presupposed that normally P ~~> Q.
    • Violated Expectation: For the utterance of Sentence1 with meaning P and Sentence2 with meaning Q to be coherent, it must be presupposed that normally P ~~> ¬Q.
    • Threat: For the utterance of Sentence1 with meaning P and Sentence2 with meaning Q to be coherent, it must be presupposed that normally ¬P ~~> Q.

    Some remaining questions I have: What about the more typical relation that or expresses, i.e., that two situations are alternatives? Would Alternatives be a kind of Contiguity relation? Why are there so few examples of non-ATB extractions with or that seem to embody this relation? The best candidate I’ve found is the pot we have to [shit] or [get off of ___], since to my mind, getting off the pot isn’t really an undesirable outcome. Also, the way it works in our house is that even if you do shit, you still eventually have to get off the pot. You might think of these as Vogon Poetry alternatives: One alternative is to get off the pot; the other is to shit … and then get off the pot.

    UPDATE, 9/15/2012: Moved first mention of the term IoD to coincide with their introduction.

5 Responses to “Songs We Have to Play or the Fans Get Angry”

  1. Glen said

    I always took “getting off the pot” to be undesirable because the sitter would prefer to stay in order to keep their options open. Someone else is telling the sitter to make a decision already; there’s an implicit ‘now’ inherent in the command: “shit [now] or get off the pot [now].”

  2. Paul Clapham said

    Here’s a strange coordination I just came across:

    “Researchers have found that the plastic stoppers so many of us use to cap an unfinished bottle, not to mention the lining of concrete vats used to store wine at many wineries, contain and can leach BPA into your glass.”

    Original source:

    I think this is an example of what you call “right-node wrapping” but I’m not sure.

  3. As you point out, the “this is the thief” examples are awkward, and I would not accept them. Part of the problem is that “this is the thief” implies you’ve already identified that thief. I decided to see if I could construct a less awkward version.

    The sentence works OK in the past tense: “This is the thief you identified, and we arrested on the spot.” (Implies, as expected, that had you not identified the thief, we would not have arrested them.)

    If there was more than one thief and the witness got a better view of some than of others, one might say: “See a thief you can identify, and we’ll arrest him on the spot.” That works, but relies on pragmatics rather than semantics for a key step in the connection (that if you can identify a thief you do). (Removing “can” gives us (“see a thief you identify…“), which I do not find acceptable because the acts of seeing and identifying are too far removed.) It would work with “recognise“, but that just makes the same pragmatic assumption in a different way.

    (Hmm … what if we replace “can” with “then” or “subsequently”? … “See a thief you then identify, and we’ll arrest him on the spot.” … maybe. Not very elegant.)

    Another approach is to make the identification conditional, with a sentence like This is the thief we hope you’ll identify, and if you do we’ll arrest him on the spot. Perfectly valid grammatically, but the extra clauses make things complicated.

    That’s all I’ve got. Do you have any more insights?

  4. (P.S. Feel free to fix my errant italics and delete this postscript.)

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: